JPS583075A - Character correcting method in recognizing device for mixture of pattern and character - Google Patents

Abstract

PURPOSE:To move a cursor quickly in a character recognizing device, by inputting a form where data are entered on a basis of a grating and separating entered patterns and characters from each other to recognize them individually. CONSTITUTION:A photoelectric converting part 1 converts photoelectrically a form, where data including pattern lines are entered, and quantizes line and the background to bidary singals and stores them in an image storage part 2. A character recognizing part 4 recognizes character images and sends recognition results to an editing part 7 together with their positions, and a pattern recognizing part 5 recognizes pattern images and sends recognition results to the editing part 7 together with their positions. The editing part 7 perform the editing for the correspondence between positions on the grating and recognition results on a basis of recognition results and positions and sends editing results to a monitor display part 8. A cursor is moved to the position of a character to be corrected, and a correct character is inputted from a keyboard part 9, and recognition results are stored in an external storage part 10 through the editing part 7.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention is based on a method for correcting corrections and misread characters in the #i recognition device, and 9. l! il
In an am device with a mixture of ll-forms and characters, which completely separates the characters and recognizes them, and edits and displays them on each NR result tube monitor, the liji-cuts and the ``l!'' of misread characters are detected. Concerning how to modify the knowledge structure.

Previously, l! with only letters! There is a method for correcting machine rejects and w4 characters in the scissor holder.

The method is as follows: ・Characters are written on a sheet of paper as shown in Figure 1 in a predetermined 9 position (placed in m and placed in the frame 0), and before starting the iII recognition , said position is stored in the recognition device as a formatting program. 0 Next, when a paper is input, the recognition device controls the position to be recognized according to said formatting program, recognizes each character in the position, and identifies each character. In this method, the recognition results are displayed on the monitor, and the characters are corrected by pointing the cursor to the H recognition result of each character and entering the correct answer from the nine-character tube keyboard.

In this case, the position where a character is written on the paper is determined by the format number, so the cursor position on the monitor and the position where one character is written on the paper are one pair.
Correspondingly, the range S in which the cursor can be moved is limited to the range in which characters can be written.In general, since a computer character display is used, the number of lines to be written on the form is large, and all lines can be displayed simultaneously on the monitor. Kl! If I can't display the characters, I scroll both screens on the monitor to slow them down so I can correct the characters.

In addition, if you display the JIO proverb result and the image of the letter O for one sheet of paper on the elegant card, you can use the keyboard to modify the proverb m conclusion without referring to the original sheet. There are ways to fix it.

However, as in the second school (O), a flowchart (flowchar) based on a grid printed in dropout color on a sheet of paper, in which both an ll shape and a letter, such as ``IllH'', are entered, is input, and the figure and letter are input. In a mixed R recognition device that separates each recognition and outputs 5 OG recognition results, the H recognition results are placed on the monitor at the position corresponding to the grid position of the paper. Even if you try to correct the characters that have been written, the position of the characters must be 7#-! Cut 1o
The fram KX* is given to the device by Ill. Also, when looking at the character writing area of the 70-chart, which is written in [1IiK] on the grid of the area scanned for wWa, characters are written only at irregular positions, so the cursor is used to correct the characters. It is not easy to move 1 grid at a time and mark the position of the character recognition result on the electrician's board.

In addition, the recognition result of a single stroke like a 70-chart shows the entire 9% fhO, and the mgm results of characters and shapes are displayed on the same screen on the monitor, and although it is round, character recognition
If you try to display l in a size that is easy to see, the size of one stroke will generally be larger than the displayable area of the monitor.
mm*I can understand your dumbfounding.

The characters are displayed too small! It becomes difficult to make corrections by looking at the iIl button.

The zero-start WA @ target inputs a good paper filled in as a grid meeting standard, separates the written shadows and characters, recognizes each k, and edits and outputs each 0g1ll conclusion. On a mixed IIA device, you can quickly move the cursor to the character-1ll definition, and it is compressed and easy to use.
You can check 1m1#II dumb, and also write sentences only on the monitor without looking at the paper! ! According to KToru/Book Ii-, which provides a character correction method that allows #O correction, (1) Input nine sheets that are filled in using the grid as a reference,
Separate the written m-shapes and letters, recognize each, and
o Recognition result tm collection In the recognition device that outputs figures and characters, the first character position of each consecutively written character string is memorized as a position on the grid. Please note the cursor movement in the cursor that was awarded in 10. A method for correcting characters that can be done from the position on the monitor that corresponds to the position on the grid. Recognized lll1Jl is enlarged and displayed at 0jlll position on the display screen (1) 0
How to correct text.

(3) Fill in the nine forms based on the grid,
Separate all the shadows and letters and teach each one,
In a mixed figure and character recognition device that outputs 01111m and 11 prefectures each, the highest @O character position of 6 character strings written consecutively in the distance between characters and figures is determined as the position on the grid. Do you remember it? K-i! Identification results OII showed 4
The cursor movement on the &ter is memorized and characters can be corrected from the O position on the electrician's tar that corresponds to the position on the circle grid.
Simultaneously move the character image pattern corresponding to the position on the grid where the cursor is moving to the position on the monitor.
! (4) The large recognition result is compressed and displayed on the monitor, and the cursor is pointing to it! The character 黴艙畢O is enlarged and displayed at the Oj1 position on the printer. How to correct the character 1 (S) Input the round paper that is written using the grid as a reference,
In a lil-form, character, and O mixed recognition device that separates the written Oma-form and the character, and edits and outputs the 011m result for each, the 011m result is edited and output. The first character position of each character string is memorized as a position on the grid, and characters can be corrected from the position on the monitor where the recognition result is displayed, and the position on the grid where the cursor is moving and its Simultaneously move multiple character image patterns corresponding to the previous IIO multiple positions to different positions on the motor! (6) The overwhelming recognition is displayed on the monitor, and the cursor points to the character "IIM" and the number of characters in front of it. (5) A method for correcting the O character, which is enlarged and displayed in a different position on the monitor, is obtained.

Below, this issue will be explained with reference to 1llIi.

jI3■ is a block diagram showing a mixed recognition device for all shapes and characters.

In the figure, 5I--sx#i party electric conversion part, 2 is image storage part, aFi character Ill shadow separation part, 4F1 character g Oribe, 5
1 is a figure recognition section, 6 is a position storage section, 7 is an editing section, 8 is a monitor/display section, 9 is a board section, and 1G is an external storage section.

As shown in No. 4611, give an incense name to each grid position on the stud,
It is assumed that it represents a written character or a projection 0 position.

Here, it is assumed that lI#'i, which represents the grid, is printed in a double-out color that is not detected by the photoelectric conversion unit l. As shown in Figure 5, the grid (indicated by broken 1 in the figure) is used as a reference. When a paper with figures and characters written on it is input, the photoelectric conversion s1 converts the character line and the aSS binary signal 111, soO is @@”K, and the converted binary signal is stored in the image memory 92. 6 shows a state 1 in which the paper shown in FIG. 5 is converted into 200 million signals by the photoelectric conversion l11 and stored in the image storage unit 2. In the figure, the binary signal 11'' is black and #@ @ ijDisplay in blank ・The top, bottom, left, and right tin O mats are also O to determine their positions. The character solid part 3 that can be associated with the honor code is stored in the image storage unit 2, and the round character and the m shape are separated, and a literary image such as 7th - is placed in the character mm part 4, and the sth kindergarten 1. Shape recognition 115K is performed on the ell engraving image, and the first position on the grid of each character string is sent to the position storage section 6.

The 12th H+ is stored in the position storage section - and indicates a good grid position O. Assuming that characters are written one by one in the grid, the round lines connected only within the grid are letters, and the wins that are connected across the grid t- Since it is a shadow, the separation part, which is the main part of the character/figure separation s3, can be constructed with a detailed block diagram of 11111Wj40, and the sensing layer is performed with the following layer thickness. Here I
In order to simplify IW 14, FIG. 15, in which the figure - is drawn so as to cross the grid m (indicated by broken marks), will be explained as a separation target.

(1) The paper containing the photoelectric conversion unit 1alllim is photoelectrically converted and quantized into a binary signal of 1- to -11 and background 1fQl, and stored in the image storage unit 2. If16m1 shows the image 111511 stored in the image storage unit 2. In II, the binary signal “1” aJl,
l (company p Expressed in blank space.

When an address is generated, the data in the image storage section 2 corresponding to that address is stored in the IllI type line wiring storage section 33. 1110 area (111117110m to h
e1@o area) is stored.

(3) The address generation unit 32 is an ILL-type collective memory 113
3c) Generate an address for accessing all locations, and output data corresponding to the number address to the concatenated address generator 34. Concatenated address generation 534Fi Graphic line memory 111s corresponding to the address generated by the address generation unit 32
If the data of 3 is -11, an address for K@mushi is issued at the 9 positions adjacent to that position (8p positions of the position where 1 point El1 matches), and the image storage unit 2 is stored at that address. The access data is accessed and set in the data pattern line storage @SS of the good image storage section 2.

(4) The address generation section 32 is the rIAs bundled storage section 33
When you return <) and access it, the figure line memory s
33 K Mori Sai (2) o Ill area (ak) K connected 111 meal S is set (1811th). This is the graphical image to be separated and is output to the compensation 41m11301, while the image storage section 2 and the type II storage section 3
The binary signal part that is different from 3 is O, which is a character image.
One side is added to the inverting input circuit 3GK, and its output is signal 8.
A character/character image signal is output as 02.

The character mm section 4 recognizes the sent character image and vi
tae*gaku is sent to the Ks elephant section 7 along with the position, ■ Kata Romaro Iku 5 recognizes the sent lll shape image tw and sends the recognition result along with the position to the editing section 7.

Mark 81B7 is based on the sent recognition information and position.
AS is performed to associate the position on the grid with the recognition result, and the resulting image is sent to the electrical worker display section 8, and the monitor display section 8 displays the edited result 1.

While looking at the 9 results displayed on the monitor display section $, use keyboard section 9 to move the cursor on the monitor, move the cursor to the character position you want to correct, enter the correct letter on keyboard section 9, and make all corrections. When the process is completed, the W iron result displayed on the monitor display section 9 is sent to the external storage section 10 through the editing section γ1 and stored in p1.

No. 911I shows an example of the display on the monitor display section 8. The Kti cursor is shown in ll, and the characters whose recognition result is rejected are ! 0 (1) Cart #K Id First, move to a fixed position on the motor, here on the monitor. (#E9-)).

(2) When a character correction O command is given to the keyboard section 9, the monitor display section 8 reads out the first character string OIk@O character position (1211th first O value SO) from the position storage section 6 and displays it on the corresponding monitor. Move the cart # to the position of The turtle cursor is at monitor 10 position O corresponding to the pulled out position.
10, and move the rigid character Oa to the first position of the character string (the position on the electrician's wheel corresponding to the 128th 3rd oam) at t.l.-(C)).

(4) Movement': If you give it to the keyboard section 9, the cursor will move one sentence at a time, so move the cursor to the reject character 0 position.
>.

(s) Look at the ILM paper on which the letters have been written and give the correct answer to the key board 9, the rejected letters will be replaced with the given correct letters), and the correction of one letter will be completed (IIIll(・)
)6 01 mentioned above! II! In the above, it was stated that only KFiNR Soba is displayed on the monitor, but if the character image is saved in the image storage unit 2 during character correction, the paper corresponding to the cursor position of the aforementioned cursor movement O11 Move the character image in the grid above to a different position on the monitor! ! (10111CI)), the characters can be easily corrected just on the M-I- without looking at the M tack where the characters were written (No. 10111CI)).

19 igm When the target characters include Noo, even if the character image of only one character is displayed, it will be distinguishable from the correct character.

In such a case, KII'i, displaying character images at a plurality of positions around 70 from the position pointed to by the cursor will make corrections much easier (1st O1ll)).

Two - I don't know much about the cabinet wag knowledge result like a chart, so I want to display letters and shapes on the monitor 10M, but I want to display the letters and shapes on the screen. When I try to display it, the general KFilEliIO dimensions are too large compared to the displayable area of the monitor, and when I try to display it with pressure, I! I can see the S-type OII recognition result, but the character recognition result 嬬/J% is displayed too quickly. When you enlarge and display the characters at a different position on the screen, you can see the previous jl! while looking at the paper on which the characters were written. Characters can be corrected using the following method (No. 1111(u)).

In addition, in the same way as mentioned above, the character at the position pointed by the cursor is enlarged and displayed at another position on the monitor, and the image memory corresponding to the position pointed by the cursor is also used!
If the character image in one lattice stored in is enlarged and displayed close to *lI11mle, the character can be easily corrected by 4 nits (Iris 1111)).

Furthermore, when the characters to be recognized by KIN include kanji,
The position pointed to by the cursor and the letter s for so*tso*toru
At the same time, character images stored in the image memory 82 in a plurality of grids corresponding to the position indicated by the cursor and a plurality of positions before and after it are enlarged and displayed If you display mgmm closest to the screen, you can easily edit text just on the monitor (
81111(e)).

This 1111th ($) to (e) O pressure line m
In the l1 stroke, you can see that it is written in 5 for ll-.

This issue @ is a combination of text and 1iII#, which is a combination of text and 1iII#, which is a stupid correction of characters on the GM device], and the general OIi image containing the text #Fai is on the monitor! When shown, remember the first position of the string and move the cursor based on that position.
Perform 1i+, enlarge the character you are writing with the cursor to a different position on the monitor, and correct the character by looking at the display tube.
It is possible for K-cycle use. (No. 1311) Also, only the characters were displayed as pressure lines on the monitor! It can also be applied to memorize the lo position, move the cursor based on the position t, enlarge the character at the position pointed by the cursor to another position, and correct the character by looking at the display. It is. (1411th) e 13111th. In FIG. 14 as well, the net-like KK thins the cursor.

As is clear from the above 011-, according to the original 11jK, #@ points are obtained for 1 as follows. (1) Only the first position of the character string is memorized, and the correction rounding is performed based on that position. With O, which can perform kernel $-, it is difficult to perform corrections quickly with a small memory capacity.

Rin) The character image at the cursor position is displayed on the monitor. Rounding, key-in, etc. are not easy 0 (!I) II
At the 9th point where a series of detailed character images are displayed, characters that cannot be uniquely determined can be corrected based on the character types and characters ** before and after.

(4) f! When the recognition results are displayed on the monitor, it is easier to enlarge the thin characters with the cursor (key-in).

[Brief explanation of drawings]

Figure 1 is an example of a form on which characters were conventionally written, Figure 2 is an example of a form on which the characters ``-'' and the ``ll'' form are present, and the 3rd WA uses the method of the present invention. Block ll0-Example, 4th block showing illumination iim device with shapes and characters
1i! Sono % 5th showing the honor mark attached to the grid on the IFi paper
The figure shows an example of letters and figures filled in based on the grid, figure 6 an example of letters and figures converted into Fit values (black and blank represent binary information), figure 7 - Figure The character image, Figure 8 is the figure image of Figure 6 @, and the 9th WJ (a) to (・) are examples of the correction procedure. An example of a monitor display when making corrections, FIG. 11 (1) to (e) is an example of a monitor display that enlarges the character at the cursor position when compressed recognition results are displayed, and shows the 6th sio + character of Daiichi Kensetsusha. The cabinet showing the first grid position of the row, Rattan 1311 and 114 are compressed and displayed on a good monitor, and the cursor position is an example of a monitor display that enlarges the character tube. The 15th cabinet is an example of writing in letters and figures, llK16 15 is a diagram in which am is binary quantized, 71117 is an example of detection of the 011 area at the grid position, FIG. 18 is a separated figure, and FIG. FIG. In the figure, 1 is a photoelectric conversion unit, 2 is an image storage unit, 3 is a character/figure separation unit, and 4 is a character ll! 5 is a figure recognition section, 6 is a position storage section, 7 is an editing section, 8 is a monitor display section,
9Fi keyboard section, 10 external storage section, K is a cursor, 31 is a grid address generation section, 32 is an address generation section, 33 is a graphic line storage section, and 14 is a follow-up address generation section. Fig. 2 1pJJ Fig. 5 Fig. 7 Fig. 8 Fig. 9 (a) wAq Fig. (b) No.'? [! 1(C) Sugar 91! I(d) #i9Figure (e) Figure 11 (a) Figure 10 (a) Figure 10 (b) Figure 11 (b) Figure 11 (v!J13 Figure 14 Figure 15 Figure 16 #, 270 Figure 18 #, Figure 19

Claims (1)

[Claims] 1. A method for combining figures and characters, which inputs a form filled out using a grid as a reference, separates and recognizes the written figures and characters, and edits and outputs each recognition result. In the character correction method for GM devices, the first character position of each character string written consecutively in the character and solid parts is memorized as a position on the grid, and the m recognition result is displayed. 4a A major correction method characterized in that character correction is facilitated by moving a cursor on the screen at a position on the monitor corresponding to the memorized position on the grid. 2. The compressed recognition results are displayed on the monitor.
The character correction method according to claim 1, wherein the character of the recognition result pointed to by the cursor is displayed enlarged at different positions on the electrical worker's table. 3. Input the friend form filled in using the grid as a reference, separate the filled 9lII form and the letters, recognize each t@, and each O
II Edit and output the knowledge result I! IjIII and letter I
& In the character correction method in a recognition device, the first character position of a string of characters written continuously is memorized as a position on a grid, and the recognition result is displayed. The cursor on the monitor is moved to a position on the monitor corresponding to the previously memorized position on the grid, and the image pattern of the character corresponding to the position on the grid to which the cursor is being moved is moved to another position on the electrician's grid. A method of correcting characters by simultaneously displaying them, checking the results, and making it easier to correct characters. 4. A compressed and nine-dimensional image is displayed on the top of the screen,
The character of the recognition result that the cursor is pointing to is
The scope of claims to be enlarged and displayed at the j1* position - the character correction method according to item 3. 5. Input the good JllI paper filled in using the grid as a reference.1
Separate the written shapes and characters, understand each kt-proverb, and edit each OII origami.l Mixture of shapes and characters.
In a character correction method for an I-weaving device, the first character position of each character string written consecutively for 1aK between the character and the ll-shape is memorized at the 0 position on the grid, and the recognition result is displayed. The cursor movement on the good monitor is performed at a position on the grid corresponding to the memorized position on the grid, and at the same time, a plurality of positions corresponding to the position on the grid where the cursor is moving and a plurality of positions before and after it are performed. The character image pattern of Kljl#KII is shown at another position on the 72-tar,
A method of correcting characters that makes it easy to correct characters without performing a check meeting and eliminates IL problems. 6. The pressure recognition result is displayed on the monitor, and the cursor is pointing to H-weave! 1lllO character and iio before it
The character correction method according to claim 5, wherein ** characters are enlarged and displayed at different positions on the monitor.